HEAD TO HEAD
Yuichi Tateno (hotchpotch): bekko-system-one and Kotoba Labs: open-jev-deberta-v3-large, compared on what they decide, where they run, what they cost and what their publishers report.
| Property | hotchpotch/bekko-system-one | kotoba-labs/open-jev-deberta-v3-large |
|---|---|---|
| Summary | Ultra-small English encoders (17M, 68M and 395M) fine-tuned from Ettin rerankers. Each answers choice, noul and score questions with a probability per candidate and no generated text. Experimental v0 by one developer; the weights have no licence assigned yet. | An independent Jev-shaped reproduction on DeBERTa-v3-large. One state and any number of choice, score and yes/no questions go in, and a calibrated distribution per question comes out of one forward pass. Public gold labels only. |
| Decides | choice, noul, score | choice, score, noul, classify |
| Architecture | bekko | open-jev |
| Fine-tuned from | cross-encoder/ettin-reranker-400m-v1 | microsoft/deberta-v3-large |
| License | Unassigned: the weights licence is not finalised (treated as non-commercial); training code MIT | apache-2.0 |
| Availability | Open weights | Open weights |
| Hosted by | — | — |
| Input price | — | — |
| Decision accuracy | — | 85.4% |
| Calibration error | — | 0.022 |
| Valid action rate | — | — |
| Median latency | 5 ms | 28 ms |
| p95 latency | 5.4 ms | — |
| Evaluation suite | The maker's short-input speed probe (4-option Choice, RTX 5090, 1 Oct 2026) | Kotoba held-out test, seen question types (banking77, SST-5, BoolQ) |
| Latest version | 2026.09 | 2026.09 |
| Variants | 0_BekkoInference | — |
| Size of latest version | 1.5 GB | 1.6 GB |
| Files | 9 | 9 |
| Downloads | 0 | 2 |
| Stars | 0 | 0 |
| Tags | system-one, encoder, ettin, onnx, browser, small, 395m | system-one, encoder, deberta, public-data, 434m |
| Updated | Oct 7, 2026 | Oct 7, 2026 |
Figures are from each model’s manifest; accuracy and latency are what the publishers report, on their own suites and hardware. Add a third model.
bekko-system-one is from Yuichi Tateno (hotchpotch) and open-jev-deberta-v3-large from Kotoba Labs. Both have open weights you can download and run. Both answer choice, noul and score questions. Only open-jev-deberta-v3-large answers classify. bekko-system-one is the smaller model, at 395M parameters to 434M. bekko-system-one is licensed other; open-jev-deberta-v3-large, apache-2.0.
Only open-jev-deberta-v3-large publishes an accuracy figure (85.4% on Kotoba held-out test, seen question types (banking77, SST-5, BoolQ)); bekko-system-one does not, so there is no comparison to make without your own test.
bekko-system-one: Free (open weights). open-jev-deberta-v3-large: Free (open weights). Open weights cost nothing per call beyond your own hardware.
By their publishers’ figures, bekko-system-one answers in about 5 ms at the median and open-jev-deberta-v3-large in about 28 ms — measured on different hardware, so treat it as a rough guide.
Yes, both: systemone pull hotchpotch/bekko-system-one and systemone pull kotoba-labs/open-jev-deberta-v3-large download the weights.